The 5 × 5 Sensor consists of emitting electrodes to establish the electric field by injecting current and probing electrodes to sense the potential perturbations. According to the DDA theory of underwater electrosense, each small emitting electrode with controlled injecting current can be seen as a point source with controlled electric charges. Thus the incident field is completely determined and controlled. The electric potential is taken between a pair of probing electrodes, rather than between the electrode and system earth, to suppress common mode noises.
To make the electric field more uniform, we can arrange the number of the pairs of the emitting electrodes and their locations. A quadrate sensing plate is designed, and each corner has a spherical emitting electrode fastened. More probing electrodes tell more information about the environmental variation, but it is hard to increase the amount because of the prototype electronics limitation (more wires and circuits). Here we design a 5 × 5 probing matrix, which will generate a 5 × 5 pixels electrical image representing the sensing region projection on such plane.
Fig. 6.1 shows the model and prototype of the electrosensor, and the func- tional diagram is shown in Fig. 6.4. The whole size of the probing matrix is 6cm × 6cm. In simulating the model using DDA approach, we only consider pure electrodes, neglecting the electrode carrier or sensor base itself. However the ac-
tual prototype is made from PCB board and wires and connectors, even though we have hollowed out several parts to reduce the effect (left one in Fig. 6.1). For applications that the body, sensor or sensor carrier (such as a robotic fish) cannot be neglected, we can use a successive reflection method [20] to model the interaction between the objects and the sensor body itself, which can be applied in DDA approach as well.
Figure 6.1: Prototypes of membrane sensor. Left figure is a two-side sensor whose electrodes are exposed to water on both side, reacting to objects from each side symmetrically. Right figure is a one-side sensor whose electrodes are insulated from water on the other side. White material is silicon sealant for protecting connectors and via holes on PCB board.
Another design is only to sense one side of the membrane (right one in Fig. 6.1), which means the electric field will not propagate to the other side. The probing electrodes are also set on one side only. Theoretically, this requires the membrane to be insulated and infinitely large, and in that case, we can simply utilize the image method to obtain the consequence that the field strength will be doubled compared to the original design. While in practical application, the field will inevitably leak to the other side. As a result, the field strength on sensing
Figure 6.2: (a)Top (red line) and bottom (blue line) view of the pcb board design of the 5 × 5 sensor. (b) Connector pin instructions.
side will be less than twice of the original design. The real fish skin is nearly a surface of a long and narrow ellipsoid, which is apparently a single side sensor.
A PCB board design is illustrated in Fig. 6.2(a). The 25 electrodes are directly lead to three connectors as pin showed in Fig. 6.2(b). These connected are further wired to a current source and a multi-input voltage meter.
As the prototype showed in Fig. 6.3, a three degree-of-freedom platform is in- house designed and manufactured for translating and rotating the sensor. Each freedom is driven by a step motor controlled with Arduino and motor driver board, which offers a serial port interface to Matlab on PC. We use commercial stimulating and measuring modules from National Instruments. These modules are fully configurable in Labview and can be seamlessly interfaced to Matlab through build-in script node. The output range of current source is ±20mA with a maximum load of 600Ω. While operating at the finest scale ±200mV of the potential meter, the absolute accuracy is 157µV and the sensitivity is 4µV .
Figure 6.3: (a) Tank and motion platform with three degrees of freedom, with experimental rubber spheres of different sizes. (b) Sensor action in the water, with stimulating and measuring electronics from Nation Instruments.
The overview of the signal processing is illustrated in Fig. 6.4. For simplic- ity, only a pair of diagonal emitting electrodes is used to inject a current signal
from stimulating source. To reject noise and reduce chemical reaction between electrodes and water, the current is modulated into a square wave at 1kHz. For this processing system, the signal is mostly digitally processed by software except for the high-speed sampling hardware. The square wave is easily generated by the program, and receiving signals on probing electrodes are synchronously de- modulated into in-phase and quadrature components. Before the demodulation, a high pass filter is utilized to removing the DC part, and the synchronous de- modulation itself acts as a super narrow band pass filter which will reject all AC components of the noise. After the demodulation, a low pass filter is set to get the final attenuation value.
Current Update _ High-pass Filter Synchronous Demodulation Low-pass Filter Sampling Sampling Attenuation Signal Generator Ref
Figure 6.4: Signal processing flowchart. The hardware works including sampling and signal update are illustrated in blue blocks, while other procedures are digitally conducted in software.
In addition to the resistant assumption of the theory, for real application, we need to consider the reactants of the water, the cables and the environment
like the tank. Also, there will be contact impedance between the electrode and the water. Further, the water impedance will change with the temperature and is hard to precisely determined without special instruments. And the geometry and electrical properties of the actual prototype will be slightly different from the simulating model. These practical influences will lead to signal mismatch between numerical simulation and experiments.
The strategy on solving these problems is by adjusting the simulation param- eters to match the electric field when no object appears. Because the electric field is uniquely determined by the potential distribution, if the simulating potentials on probing electrodes are matched to the experimental ones, then the electric field is believed to be matched. Consequently, perturbation signals of the object, which are used in the sensing algorithms, are believed to be consistent.
More specifically, two parameters including the injecting current and the water conductivity are referred. The experimental and simulating current are set to be the same, but the water conductivity is a variable in the simulation. After the experimental potentials are collected, we run a series of simulations to search for the best value of the simulating water conductivity, which will minimize the mean square error between experimental and simulating potentials. After that, the perturbation signals are also matched, and we can use these parameters in sensing algorithms.